Agents UI

Expand

Agents list (/agents)

The Agents page shows all configured agents as cards with:

• Avatar and name
• Archetype badge
• Model and provider
• Last active timestamp
• Marketplace badge (for agents installed from the marketplace)

Click an agent card to open that agent in chat. Use the edit button to open the workspace editor. Click New Agent to start the Awakening experience.

Agents can be deleted with a slide-to-delete gesture on their card.

Awakening — creating an agent

The Awakening experience is a cinematic onboarding flow for creating a new agent. Clicking New Agent redirects to /awakening?force=true, which loads a full-screen 3D experience (served as static files).

A quick awakening variant is available at /awakening/quick for faster agent creation without the full cinematic sequence.

1. Welcome

   An animated orb pulses on screen with the heading “Awakening…” and a brief introduction. Click Begin to start creating your agent, or Skip to use defaults.

2. Name your agent

   Enter a display name for your new Snippbot and optionally provide your own name. The name is checked for availability in real time.

3. Choose an avatar

   Pick or upload an avatar image for the agent using the avatar picker.

4. Choose an archetype

   Twelve archetypes appear as animated cards. You can select up to three to create a blended personality. Click to toggle selection.

   Archetype	Description
   Personal Companion	Closest digital friend, confidant, and emotional support
   Professional Assistant	Executive-level secretary, organized, anticipates needs
   Researcher	Deep diver, curious explorer, knowledge synthesizer
   Teacher / Mentor	Patient guide, explains concepts, grows understanding
   Creative Partner	Brainstormer, muse, collaborator on artistic endeavors
   Coach / Motivator	Accountability, motivation, pushing toward goals
   Technical Expert	Deep technical knowledge, code review, architecture
   Entrepreneur Partner	Business strategist, idea validator, MVP builder
   Health / Wellness	Supportive presence for health, habits, mindfulness
   Accountability Partner	Keeps you on track, follows up, celebrates wins
   Thinking Partner	Think through complex problems, brainstorm solutions
   Something Unique	Describe exactly who you need (custom free-text)

   If you select “Something Unique,” a text field appears to describe your custom archetype.

5. Review profile

   A generated personality profile shows editable attributes (Core Drive, Communication Style, Strength, Approach, Values) and a summary sentence. You can edit any attribute inline or regenerate the profile.

6. Configure voice

   Select a TTS provider and voice for the agent. You can preview voices before confirming.

7. Getting-to-know-you questions

   The agent asks three personalized questions generated from your chosen archetype(s). Answer each one to help the agent understand how to work with you. Each answer is followed by a short reflection from the agent.

8. Evolution preference

   Choose how the agent should grow over time:

   • Evolve freely — grow and change based on your journey together
   • Stay focused — remain specialized in this role
   • Check with me — ask before making significant changes

9. Complete

   A summary screen shows the agent’s name, archetype(s), and a greeting. Click Begin our journey to create the agent and open it in chat.

Asking an agent to create one

You can also ask an agent in chat to create a new agent. For example: “Make me a persistent market-research agent that focuses on pricing.” The agent uses its built-in create_agent tool:

1. It asks you first. The agent shows the proposed name, role and description in a confirmation question. Nothing is created until you approve.
2. It creates the agent the same way Awakening does. The new agent gets its own workspace (IDENTITY.md, USER.md, NOTES.md, IDLE.md, VOICE.json) and an @handle. It appears in the agents list and the room’s Add agent menu straight away. Its USER.md is copied from the agent that created it, so it already knows who you are.
3. It doesn’t join the room automatically. To bring the new agent into the current conversation, the agent asks again and then adds it. After that it shows up in the @ picker like any other room member.

When you make the request, you can also ask for:

Option	Effect
Expertise tags	Used when agents look for a specialist to add to a room
Archetypes (up to 3)	Set the agent’s guiding philosophy, as in Awakening
Evolution	Check with me (default), Stay focused, or Evolve freely
Role lock	Adds a section to IDENTITY.md that keeps the agent strictly in its domain
Starting notes	Seeds NOTES.md, for example with what a sub-agent learned
Model	The agent’s preferred model, for example claude-native:sonnet
Tool focus	Narrows which tools the agent is offered (core tools always stay). This is a focus, not a security boundary.
Avatar	A profile picture for the agent. The confirmation question also asks whether you want one.

Avatars

An agent can give any agent a profile picture with the built-in agent_avatar tool. Ask for it while creating the agent, or later: “Make Rita an avatar: a friendly analyst with round glasses.”

• It uses your ChatGPT subscription by default, at no extra cost. If that isn’t available, the agent tells you why. It never switches to a paid generator on its own.
• Gemini costs money. Gemini charges per image, so the agent first tells you the estimated cost and asks you to approve it.
• Any other generator works too. The agent makes the image with that tool (for example a fal tool, after you approve the cost), then attaches the file.
• It’s saved where an upload would be. The picture is cut to a square, stored at 512 px, and saved to the same place as an avatar you upload on the agent’s edit page. It shows everywhere that agent appears.
• It asks before replacing an avatar. Each agent has one avatar, so a new one overwrites the old one. If the agent already has an avatar, the tool won’t replace it until you say so.

Sub-agents, scheduled jobs and Plan Mode can’t create agents. To remove an agent, delete it from the agents list.

Agent workspace editor (/agents/:id/edit)

The edit page provides a two-panel layout for editing agent workspace files:

File sidebar (left)

Lists the editable workspace files that define the agent’s personality and behavior:

File	Purpose	Writer
IDENTITY.md	Who the agent is — persona, role, voice, core behaviors. Frozen during a session for cache stability.	You; agent via identity_patch (gated by Evolution Mode)
USER.md	What the agent knows about you — preferences, projects, communication style.	You; agent via user_note
NOTES.md	The agent’s running observations — conventions, discoveries, what worked.	You; agent via memory_note (Phase 7 reflection writes here too)
IDLE.md	Proactive behavior directives — when and how the agent reaches out.	You

If you migrated an agent from the v1 workspace (PERSONA/PROFILE/BEHAVIORS/SKILLS), the originals are preserved in _archive_v1/ and can be browsed from the editor’s Archived (v1) section in the sidebar — read-only, so you can copy wording forward without losing the source of truth.

Editor (right)

Click any file to edit it in the right panel:

• Edit/Preview toggle — switch between raw markdown and rendered preview
• Save with Cmd+S shortcut or the Save button
• Manual edits take effect immediately. Agent-written memory_note / user_note changes take effect from the agent’s next reply (the agent’s context is rebuilt every turn — except Claude Code (claude-native) models, whose resumed session keeps the context it started with, so for them it’s the next conversation); identity_patch changes follow the agent’s evolution mode

Avatar and voice

At the top of the edit page you can:

• Click the avatar to upload a new image (with crop modal)
• Configure the voice settings (TTS provider and voice selection)
• Delete the agent (with confirmation)